Deep Representation Learning in Speech Processing: Challenges, Recent Advances, and Future Trends

Abstract

Research on speech processing has traditionally considered the task of designing hand-engineered acoustic features (feature engineering) as a separate distinct problem from the task of designing efficient machine learning (ML) models to make prediction and classification decisions. There are two main drawbacks to this approach: firstly, the feature engineering being manual is cumbersome and requires human knowledge; and secondly, the designed features might not be best for the objective at hand. This has motivated the adoption of a recent trend in speech community towards utilisation of representation learning techniques, which can learn an intermediate representation of the input signal automatically that better suits the task at hand and hence lead to improved performance. The significance of representation learning has increased with advances in deep learning (DL), where the representations are more useful and less dependent on human knowledge, making it very conducive for tasks like classification, prediction, etc. The main contribution of this paper is to present an up-to-date and comprehensive survey on different techniques of speech representation learning by bringing together the scattered research across three distinct research areas including Automatic Speech Recognition (ASR), Speaker Recognition (SR), and Speaker Emotion Recognition (SER). Recent reviews in speech have been conducted for ASR, SR, and SER, however, none of these has focused on the representation learning from speech -- a gap that our survey aims to bridge.

arXiv commentPart of this work is accepted in IEEE Transactions on Affective Computing 2021. https://ieeexplore.ieee.org/document/9543566

References in corpus (57)

• Sequence to Sequence Learning with Neural Networks
• Supplementary information for "Quantum supremacy using a programmable superconducting processor"
• Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks
• A Brief Survey of Deep Reinforcement Learning
• WaveNet: A Generative Model for Raw Audio
• Neural Discrete Representation Learning
• VoxCeleb2: Deep Speaker Recognition
• Conditional Image Generation with PixelCNN Decoders
• Pixel Recurrent Neural Networks
• Training Very Deep Networks
• Theoretical Models of Learning to Learn
• Exact solutions to the nonlinear dynamics of learning in deep linear neural networks
• Deep Learning for Audio Signal Processing
• Adversarial Feature Learning
• MelGAN: Generative Adversarial Networks for Conditional Waveform Synthesis
• Deep Speaker: an End-to-End Neural Speaker Embedding System
• Deep Reinforcement Learning for Dialogue Generation
• Recent Advances in Autoencoder-Based Representation Learning
• InfoVAE: Information Maximizing Variational Autoencoders
• Deep Variational Information Bottleneck
• Deep Generative Stochastic Networks Trainable by Backprop
• Towards a Definition of Disentangled Representations
• Conditional Generative Adversarial Networks for Speech Enhancement and Noise-Robust Speaker Verification
• GANSynth: Adversarial Neural Audio Synthesis
• Did you hear that? Adversarial Examples Against Automatic Speech Recognition
• THCHS-30 : A Free Chinese Speech Corpus
• Domain Adversarial for Acoustic Emotion Recognition
• Self-supervised Visual Feature Learning with Deep Neural Networks: A Survey
• Feature Learning in Deep Neural Networks - Studies on Speech Recognition Tasks
• k-Sparse Autoencoders
• Learning Hidden Unit Contributions for Unsupervised Acoustic Model Adaptation
• Semi-Supervised Speech Emotion Recognition with Ladder Networks
• Speaker-Invariant Training via Adversarial Learning
• Neural Models for Information Retrieval
• ESPnet: End-to-End Speech Processing Toolkit
• DeepMDP: Learning Continuous Latent Space Models for Representation Learning
• Invariant Representations for Noisy Speech Recognition
• Generative adversarial network-based glottal waveform model for statistical parametric speech synthesis
• Active Learning for Speech Recognition: the Power of Gradients
• An Unsupervised Autoregressive Model for Speech Representation Learning
• Disentangling the independently controllable factors of variation by interacting with the world
• Hierarchical Generative Modeling for Controllable Speech Synthesis
• Adversarial Machine Learning And Speech Emotion Recognition: Utilizing Generative Adversarial Networks For Robustness
• Adversarial Speaker Adaptation
• Exploiting Low-dimensional Structures to Enhance DNN Based Acoustic Modeling in Speech Recognition
• Adversarial Network Bottleneck Features for Noise Robust Speaker Verification
• Towards adversarial learning of speaker-invariant representation for speech emotion recognition
• Towards Speech Emotion Recognition "in the wild" using Aggregated Corpora and Deep Multi-Task Learning
• Learning Representations of Affect from Speech
• Applying Cooperative Machine Learning to Speed Up the Annotation of Social Signals in Large Multi-modal Corpora
• Unsupervised Domain Adaptation by Adversarial Learning for Robust Speech Recognition
• Abstract Learning via Demodulation in a Deep Neural Network
• N-HANS: Introducing the Augsburg Neuro-Holistic Audio-eNhancement System
• A study of semi-supervised speaker diarization system using gan mixture model
• Towards Learning Fine-Grained Disentangled Representations from Speech
• Autoencoder based Domain Adaptation for Speaker Recognition under Insufficient Channel Information
• Abstraction Learning

Cited by in corpus (11)

• Self-Supervised Speech Representation Learning: A Review
• Towards Learning a Universal Non-Semantic Representation of Speech
• Viewmaker Networks: Learning Views for Unsupervised Representation Learning
• A novel policy for pre-trained Deep Reinforcement Learning for Speech Emotion Recognition
• Seen and Unseen emotional style transfer for voice conversion with a new emotional speech dataset
• Deep Reinforcement Learning with Pre-training for Time-efficient Training of Automatic Speech Recognition
• Emotional Voice Conversion: Theory, Databases and ESD
• End-to-end Recurrent Denoising Autoencoder Embeddings for Speaker Identification
• It's not what you said, it's how you said it: discriminative perception of speech as a multichannel communication system
• Expressive Voice Conversion: A Joint Framework for Speaker Identity and Emotional Style Transfer
• Speech Representations and Phoneme Classification for Preserving the Endangered Language of Ladin